Undergraduate Upends a 40-Year-Old Data Science Conjecture

A 1985 paperComputer scientist Andrew YaoThe KAM Turing Award will win, he sees the serted that the best way to find a separate element or empty spot in the hash tables with a specific set of a particular property is to go through randomly possible spots – this is an approach known as the identical probing. He also said that in the worst situation, where you are searching the last open space, you can never do better than that XThe For 40 years, most computer scientists assumed that Yao’s estimate was true.

He was unknown about this for common reasons that Crappivin was not kept behind by conventional knowledge. “I did it without knowing Yao’s guess,” he said. His inquiries with small pointers led to a new type of hash table – it is a one that does not depend on the same investigation. And for this new hash table, the time required for the worst case search and insertion is proportional (log X)²– faster than XThe This result directly opposed Yao’s hypothesis. Helped to show it to Pharach-Kom-Corton and Kusjamul Crappivin (log) X)² Yao is the best, unbeatable bound for the popular class that Yao wrote.

“This result is beautiful that it adds to this classic problem and resolves,” says Guy blillok Carnegie Mellon’s.

“It’s not just the one that they refused [Yao’s conjecture]They also found the best possible answers to his question, “said Sepa Waterloo University. “Before we know the correct answer we could go 40 years.”

Craopivin on the King College Bridge at Cambridge University. His new hash table can find and store data faster than the way researchers thought it was possible.

Photoraf: Philip Amone for Quanta Magazine

In addition to refuting Yao’s estimates, many of the new paper also consider more wonderful results. Related to it, though somewhat different, in the situation: in 1985, Yao saw not only in the worst time for questions, but also on the average time covered by all possible questions. He proved that “greedy” labeled some specific characterized hash tables – which means new elements must be placed on the first available spot – the average time than the log can never achieve better. XThe

Pharach-Corton, Crapvin and Kusjamul wanted to see if the same limit was applied to non-garde hash tables. They showed that it does not provide a counter example, a non-garde hash table with an average queries that is much better than log XThe In fact, it does not depend on X At all “You get a number,” Pharach-Kom-Corton said, “something that does not depend on just how full a constant and the hash table.” It was completely unexpected that you could achieve a constant average query time regardless of the fullness of the hash table – even to writers.

The team’s results cannot lead to any immediate appeal, but it is not all the issues, said Conway. “This type of data structure is more important to understand better. You don’t know when the results like this will unlock something that lets you do better in practice “”

Real story Re -printed with permission How many magazinesAn editorially independent publishing Simon’s Foundation Whose aim is to increase the public understanding of the science of mathematics and the development of the research of physical and life science and covering the trends.